Enhancement request: Allow exclusion based on adjusted r^2

[Andy00000000000]

Hi,

Really great package that makes NCA quite easy, thank you for that contribution!

I was wondering if there is an option for implementing lambda z/half-life exclusion from summary statistic tables using adjusted r2 (i.e., $min.hl.r.squared but for adj.r.squared)?

Thanks,

Andy

[billdenney]

Currently, the exclusion is only setup based on r-squared and not adjusted. It's a good suggestion to add, so I will do so.

[Andy00000000000]

Great, looking forward to it, thanks!

As a short-term workaround (which appears to be working as intended after manually checking) I copied the exclude_nca_min.hl.r.squared function and modified to pull adj.r.squared while still utilizing the min.hl.r.squared option; then passed the modified function into the exclude() call.

exclude_nca_min.hl.adj.r.squared <- function (min.hl.r.squared){
  affected_parameters <- get.parameter.deps("half.life")
  missing_min.hl.r.squared <- missing(min.hl.r.squared)
  function(x, ...) {
    if (missing_min.hl.r.squared) {
      min.hl.r.squared <- PKNCA.options("min.hl.r.squared")
    }
    ret <- rep(NA_character_, nrow(x))
    if (!is.na(min.hl.r.squared)) {
      idx_r.squared <- which(x$PPTESTCD %in% "adj.r.squared")
      if (length(idx_r.squared) == 0) {
      }
      else if (length(idx_r.squared) == 1) {
        current_r.squared <- x$PPORRES[idx_r.squared]
        drop_r.squared <- !is.na(current_r.squared) &
          current_r.squared < min.hl.r.squared
        if (drop_r.squared) {
          ret[x$PPTESTCD %in% affected_parameters] <- sprintf("Half-life adj.r-squared < %g",
                                                              min.hl.r.squared)
        }
      }
      else if (length(idx_r.squared) > 1) {
        stop("Should not see more than one r.squared (please report this as a bug)")
      }
    }
    ret
  }
}

o_result1 <- exclude(o_result1,FUN = exclude_nca_min.hl.adj.r.squared())

[billdenney]

Yep, that's a perfect work-around. The final solution within the package will look very similar to that (with more documentation, tests, etc.).